Photo: Shutterstock
Artificial intelligence is transforming medicine, but what does that mean for breast cancer screening? Researchers at UMass Chan Medical School are exploring how an AI-based tool might help identify women at higher risk for breast cancer and in doing so, detect cancers that standard mammograms might miss.
According to the American Cancer Society, breast cancer remains one of the most common cancers among women in the United States, and early detection is key to improving outcomes. Mammography, the current standard screening test, saves lives but has limitations, especially for women with dense breast tissue, where tumors can be harder to see.
Mohammed Salman Shazeeb, PhD, associate professor of radiology, and Gopal Vijayaraghavan, MD, MPH, professor of radiology, are part of a team testing an AI-driven risk assessment model developed in collaboration with investigators at the Massachusetts Institute of Technology. Supported by grants from state agencies and the Breast Cancer Research Foundation, the tool analyzes routine screening mammograms and assigns a risk score that estimates a woman’s likelihood of developing breast cancer in the next few years.
A targeted approach to supplemental screening
Instead of recommending supplemental imaging for every patient, the AI risk score helps the study team identify a smaller cohort of women who may benefit most from additional testing.
“Among the roughly 6 to 7 percent of women who scored above our risk threshold, we invited them for contrast-enhanced breast MRI,” said Dr. Shazeeb. “What’s striking is that all had normal screening mammograms, yet MRI found cancers in some of them that we would otherwise have missed.”
In the first 145 study participants, four additional cancers were identified through MRI despite negative mammography results—a yield that is several times higher than what a mammography screening alone typically detects in a similar number of women.
“The tool is trained for performance, not understanding. That’s why these tools are designed to augment, not replace, clinical judgment.”
“MRI remains the gold standard for detecting many breast cancers,” said Dr. Vijayaraghavan. “But it’s expensive, time-consuming and not feasible for everyone annually. A tool that helps us focus those resources on women at highest risk could make early detection more efficient and more personalized.”
How AI “sees” risk
AI models can detect imaging features too subtle for the human eye, drawing on patterns learned from large datasets. This expansive pattern recognition is one reason they can reveal risk signals that clinicians might miss.
“The AI can process many more features on an image than a radiologist can visually,” Vijayaraghavan explained. “But it doesn’t think the way a physician does. The tool is trained for performance, not understanding. That’s why these tools are designed to augment, not replace, clinical judgment.”
Human oversight remains essential, both to interpret what the AI highlights and to ensure that the tool doesn’t flag insignificant findings. As researchers continue to study this balance between sensitivity and specificity, larger studies will be needed to fully validate the tool across populations.
Challenges and future steps
While the early findings are promising, the path to clinical use involves several hurdles. FDA approval, reimbursement policies and ensuring equitable access are all questions that must be addressed before AI-guided risk assessment becomes routine in clinical care.
“Before this can be widely applied, we need larger-scale validation and real-world implementation data,” Shazeeb said. “We also have to ensure that it works fairly across diverse populations and different mammography systems.”
Another part of the research effort is patient engagement. While patient responses to the role of AI in imaging is generally mixed, Sara Schiller, senior research program manager in the Department of Radiology, said, “Many women I speak with just want to help further research and are not hesitant about AI per se. Many have family histories of breast cancer and are eager to contribute.”
A personalized future for screening
Experts emphasize that AI is not a magic solution, but a potential tool that could help tailor screening to individual risk.
“The goal isn’t to replace mammograms,” said Vijayaraghavan. “It’s to add another layer of insight, a decision support tool, that helps us find cancers earlier, when treatment is more effective and less invasive.”
As research continues, tools such as this may help usher in a more personalized approach to breast cancer screening, one that aligns the latest technology with clinical expertise to improve outcomes for women.